Surface Reactions During Silicon Oxide Etching Using Fluorocarbon Plasma
Kenji Ishikawa
Abstract
Fabrication of ultra-large-scale integrated (ULSI) circuits using silicon metal-oxide semiconductor devices requires that costs must be lowered, the process design cycle must be shortened, and the manufacturing must be controlled. In the ULSI fabrication, “Plasma etching” is a key process because of the only commercially viable technology for anisotropic removal of material from surfaces. The process can be chemically-selective, removing one type of material from surfaces while leaving other materials unaffected, and can be anisotropic, removing material at the bottom of a trench while leaving the same materials on the sidewalls unaffected. As the dimensions of next-generation ULSI circuits continue to shrink, design and control of the plasma etching processes are becoming more and more challenging with respect to achieving anisotropy with high selectivity. To realize these objectives, we must fully understand the chemical reaction mechanism during processing, both chemo-physically and quantitatively.
Fluorocarbon gas plasmas are used as standard gas chemistries in silicon dioxide (SiO2) etching processes. In the fluorocarbon plasma process, a film of amorphous fluorinated carbon (a-C:F) deposit on the surface and render both sidewalls and bottom of a trench, which are unreactive in absence of sufficient ion flux and insufficient radical flux. During the etching of SiO2, the a-C:F is consumed by complex ion-assisted reactions represented by:
SiOs +CFx(g) → SiOCFs +M+(g) → COx ↑ +SiFx ↑ +M,
where s and g stand for surface and gas species, and M+ is any ion. Thus, it is considered that the a-C:F film (SiOCF in the above equation) plays an important role on the etching performance. Therefore, to characterize the formation of the a-C:F film as part of gaining a full understanding of the etching reaction mechanism is needed.
To clarify the reaction mechanism, the author focuses on characterization of changes in the structural properties of both SiO2 and a-C:F films during the etching. This is desirable a non-destructive, non-contact analysis method. For this purpose, an in-situ infrared (IR) spectroscopic measurement was chosen in this study.
First, IR spectra are consisted of a summation of optical responses for components of materials. To discriminate each components by analyzing the IR spectra, structural and optical properties of individual components, such as SiO2 and a-C:F, are needed. Therefore, sets of the dielectric function $\epsilon(\omega)$ for the films were first prepared. The $\epsilon(\omega)$ spectra were calculated by requiring the calculated spectra to fit actual spectra using the spectra simulation method, being developed by the author. The actual peak line shape of the imaginary part ($\epsilon "$) of the dielectric function is described by a Gaussian, but not by a Lorentzian. More specifically, high– and low–frequency edges at half–height for the imaginary part of the dielectric function are not actually symmetrical. The author thus developed a dielectric function model that describes an asymmetrical Gaussian line shape. This model reduces the number of parameters needed to the center position distribution width and the low– and high–Gaussian distribution widths, fewer than the number needed for the assignment method for multiple purely Gaussian peaks. As the result, the measured infrared spectra can be precisely analyzed.
Next, to control the performance of the plasma etching process, the author conducted in-situ real-time IR spectroscopy measurements. To gain sensitivity for ultrathin films, an attenuated-total-reflection (ATR) method was used. The IR-ATR spectra exhibit discriminated absorptions of the bands of carbon-fluorine and silicon oxygen. Using a biased substrate, the author took a series of spectra every half-second during the etching and obtained real-time intensity profiles by decomposing the spectra into absorption attributed by the CF and SiO bonds. During the etching, the a-C:F film formed on the surface, and its thickness immediately reached a steady state. The author used a model based on the balance between the rates of deposition and sputtering of a-C:F film to investigate the time dependence of the film formation. In Ar-diluted c-C4F8 plasma, the steady-state thickness of the film was about one fifth that on the surface of silicon (0.5 vs. 3 nm). Thus the formation of a-C:F film on silicon-oxide surface during its etching in fluorocarbon plasma was observed. Specifically, changes in the surface during the early stages were quantitatively characterized, and the differences between surfaces of silicon-oxide and silicon were identified. These results indicate that the rate of a-C:F film consumption was roughly comparable to the rate of oxide etching.
The chemistry associated with fluorocarbon plasmas is complex because of the complicated interactions between the gas-phase radicals and the surface; more work is needed to enable correlation between the density of gas-phase radicals and the thickness of a-C:F film on a polymerized surface. The relationship between the density of CFx radicals in the gas phase and the condition of a polymerized surface, such as the thickness of a-C:F film on the surface, needs to be clarified.
To realize this requirement, the author combines in-situ IR spectroscopic characterization of the surface and spatially resolved laser-induced fluorescence (LIF) observation of the density of the gas-phase radicals using planar LIF measurements. The author thus measured the density of CF2 radicals in inductively coupled plasmas (ICP) generated using an Ar-diluted c-C4F8 gas mixture. The relationship between the axial profiles of the density of the radicals in the gas phase and the thickness of the a-C:F film was examined using time-resolved simultaneous planar LIF measurements and IR spectroscopy.
In the background, several models have been proposed to explain the interaction between gas-phase radicals and surface. Following the models, the generation and loss of CFx radicals in the gas phase have been discussed. (1) CFx radicals generated as a result of ion irradiation of a polymerized surface circulate between the gas-phase and the surface and are destroyed by oligomerization in the gas phase (“the cyclic model”). (2) CFx+ ions are converted into neutral radicals at the surface (“the reflection model”). (3) CFx radicals are lost by becoming adsorbed on the surface when it is irradiated by highly energetic particles such as ions (“the adsorption model”). (4) CFx radicals are generated in the plasma-sheath region as a result of collisions with highly energetic ions (“the sheath-generation model”). (5) The density of CFx radicals in plasma becomes sparse due to a significant increase in the temperature (“the sparse model”). (6) CFx radicals are destroyed as a result of electron-impact-induced dissociation in the bulk plasma (“the destruction model”). To clarify the models, the author carried out in-situ time-resolved simultaneous measurements of the density of gas-phase radicals and observations of the surface.
The relationship between the axial profiles of the density of the radicals in the gas phase and the thickness of the a-C:F film was clarified and the experiments revealed it that relatively few CF2 radicals are generated on a polymerized surface and that the main determinant of CF2 radical density is likely the electron-impact-induced destruction of the CF2 radicals themselves. Therefore, gas-phase radical determination is mainly followed by the destruction model. Especially, in high density plasmas generated by such ICP, surface modification becomes more important in the chemical reactions.
Reactive species such as ions impinging on the surface accelerate due to the biasing substrate. Surface chemical reactions thus occur, forming products that are desorbed. To clarify these elemental reactions, the author used an ion-beam apparatus to investigate the interactions between the SiO2 surface and the incident species. In actual etching processes with Ar-diluted c-C4F8 plasma, half of the incident ions on the surface are carbon mono-fluoride (CF+) with accelerating energies ranging from eV to keV. Thus, the author investigated particularly the interaction of the CF+ ions.
Previous beam studies showed that early-stage surface modification, for instance, of a-C:F film occurs when a neutral species such as CF2 is deposited. However, the contribution of single ions to the surface modification when the neutrals are eliminated remains unclear. Two factors in particular need to be investigated: (1) the contribution of a single CF+ ion with an incident energy of a few hundred eV to the deposition of a-C:F film and (2) the change in surface conditions as the ion dose is increased.
The author investigated surface of SiO2 film after irradiated by mass-analyzed fluorocarbon ion beam with an incident energy below 1 keV. Both etching yields for individual ion and early-stage surface modification depended on the ion beam dosage. Notably, experimental results showed that the regimes from etching to deposition changed transitionally. At a dose of less than 1.0×1017 cm−2, SiO2 etching was observed. At a dose greater than 1.0×1017 cm−2, a-C:F film was continuously deposited on the modified surface, on which carbon accumulated. The contribution of surface conditions was clearly observed using in-situ x-ray photo-electron spectroscopy analysis.
Finally, the author developed an in vacuo electron-spin-resonance (ESR) technique to investigate the chemical structure with respect to dangling bonds. Such in vacuo ESR measurements thus provide a new experimental approach to the microscopic understanding of chemical reactions that affect etching processes. In the a-C:F films deposited by the fluorocarbon plasmas, the C dangling bonds (DB) density of a-C:F films is quite high (2×1021 cm−3) and stable in vacuum at room temperature. The author also found that the C-DB had a high chemical reactivity with oxygen molecules. From these results, the author tentatively explained the effect of a-C:F film on SiO2 etching done using a fluorocarbon gas system. Moreover, the author investigated Si-DB created during plasma etching of SiO2 using an Ar-diluted fluorocarbon gas mixture and the in vacuo ESR. The Si-DB in SiO2, namely E' center, was mainly created about 10 nm beneath the a-C:F film due to irradiation of vacuum ultraviolet rays from the plasma. These experimental results suggest that the interaction between a top-covered a-C:F layer and a defective SiO2 layer underneath it accelerates the etching rate.
In summary, the author have investigated the chemical reactions during plasma etching of SiO2 using fluorocarbon plasmas. (1) Combining the simultaneous planar LIF measurements with the IR spectroscopy, interactions between the gas-phase radicals and the surface in the high-density ICP were revealed that the main determinant of gaseous CF2 radical density was likely the electron-impact-induced destruction of the CF2 radicals themselves. (2) Interactions between the SiO2 surface and the mass-analyzed fluorocarbon ions (CF+) were sensitive to the surface condition, depending on the ion dosage. The characteristic sequence of transitional changes in etching and deposition was studied. (3) Using the in-situ real-time IR spectroscopy measurements, the in-situ experiments during the etching were revealed that the a-C:F film formed on the surface, and its thickness immediately reached a steady state. (4) The in vacuo electron-spin-resonance (ESR) have been developed to investigate both C-DB and Si-DB on the surface during plasma etching. The C-DB density in the a-C:F film is quite high the order of 1021 cm−3. The C-DB are chemically reactivity with oxygen.
Above mentioned, the author could clarify chemo-physically the surface reactions during etching of SiO2 using fluorocarbon plasmas.